I've been trying to streamline the Linux NTP support in the last
months, but didn't dare to change too much. With the message
attached, I suspect we should work on standardized NTP support for
the kernel if we don't want to drop it completely. Especially the
last sentence worries me...
Ulrich
------- Forwarded Message Follows -------
Date: Fri, 20 Sep 96 13:42:23 EDT
From: Mills@huey.udel.edu
To: Ulrich Windl <ulrich.windl@rz.uni-regensburg.de>
Cc: Mills@huey.udel.edu, time@informatik.uni-erlangen.de
Subject: Re: xntp3-5.85: Linux trouble
Ulrich,
I get about ten messages a day on some problem with NTP, some of them
pretty green. Of those, eight are on Linux and of those four are
on the darn timex.h problem. I've hollered and screamed for two years
on this problem, but the Linux folk seem not to listen. It's real
easy to fix, just use the current timex.h header file in the
current kernel.tar.Z release, which has been stable for two years.
My present course is to disable the kernel modifications, unless some
interested soul cares to unravel the fine points. As it is, half the
modifications (PPS support) don't work anyway with the broken
structure definitions. The FreeBSD version works just fine. My marbles
are on that system.
Dave